请问用到了GPU加速吗

[xiatian815]

Args in experiment:
Namespace(activation='gelu', affine=0, batch_size=128, c_out=7, checkpoints='./checkpoints/', d_ff=256, d_layers=1, d_model=128, data='custom', data_path='weather.csv', dec_in=7, decomposition=0, des='Exp', devices='0,1,2,3', distil=True, do_predict=False, dropout=0.2, e_layers=3, embed='timeF', embed_type=0, enc_in=21, factor=1, fc_dropout=0.2, features='M', freq='h', gpu=0, head_dropout=0.0, individual=0, is_training=1, itr=1, kernel_size=25, label_len=48, learning_rate=0.0001, loss='mse', lradj='type3', model='PatchTST', model_id='336_96', moving_avg=25, n_heads=16, num_workers=10, output_attention=False, padding_patch='end', patch_len=16, patience=20, pct_start=0.3, pred_len=96, random_seed=2021, revin=1, root_path='./dataset/', seq_len=336, stride=8, subtract_last=0, target='OT', test_flop=False, train_epochs=100, use_amp=False, use_gpu=True, use_multi_gpu=False)
Use GPU: cuda:0

start training : 336_96_PatchTST_custom_ftM_sl336_ll48_pl96_dm128_nh16_el3_dl1_df256_fc1_ebtimeF_dtTrue_Exp_0>>>>>>>>>>>>>>>>>>>>>>>>>>
train 36456
val 5175
test 10444
iters: 100, epoch: 1 | loss: 0.6968859
speed: 0.8145s/iter; left time: 23051.8019s
iters: 200, epoch: 1 | loss: 0.6983635
speed: 0.6318s/iter; left time: 17817.0645s
Epoch: 1 cost time: 198.10548210144043
Epoch: 1, Steps: 284 | Train Loss: 0.7457502 Vali Loss: 0.5418562 Test Loss: 0.2217388
Validation loss decreased (inf --> 0.541856). Saving model ...
Updating learning rate to 0.0001
iters: 100, epoch: 2 | loss: 0.3526905
speed: 1.9244s/iter; left time: 53915.4483s
iters: 200, epoch: 2 | loss: 0.4079723
speed: 0.6331s/iter; left time: 17674.2574s

[lileishitou]

how to use learner.distributed(), in self supervised pretrain code ？

[WHU-EE]

I have the same question, have you solve it?
my gpu is rtx 3060
this is partial log:
iters: 100, epoch: 1 | loss: 1.0063255
speed: 0.7884s/iter; left time: 22311.7481s
iters: 200, epoch: 1 | loss: 0.6996360
speed: 0.3332s/iter; left time: 9397.4480s
Epoch: 1 cost time: 141.5194652080536

[WHU-EE]

Args in experiment: Namespace(activation='gelu', affine=0, batch_size=128, c_out=7, checkpoints='./checkpoints/', d_ff=256, d_layers=1, d_model=128, data='custom', data_path='weather.csv', dec_in=7, decomposition=0, des='Exp', devices='0,1,2,3', distil=True, do_predict=False, dropout=0.2, e_layers=3, embed='timeF', embed_type=0, enc_in=21, factor=1, fc_dropout=0.2, features='M', freq='h', gpu=0, head_dropout=0.0, individual=0, is_training=1, itr=1, kernel_size=25, label_len=48, learning_rate=0.0001, loss='mse', lradj='type3', model='PatchTST', model_id='336_96', moving_avg=25, n_heads=16, num_workers=10, output_attention=False, padding_patch='end', patch_len=16, patience=20, pct_start=0.3, pred_len=96, random_seed=2021, revin=1, root_path='./dataset/', seq_len=336, stride=8, subtract_last=0, target='OT', test_flop=False, train_epochs=100, use_amp=False, use_gpu=True, use_multi_gpu=False) Use GPU: cuda:0

start training : 336_96_PatchTST_custom_ftM_sl336_ll48_pl96_dm128_nh16_el3_dl1_df256_fc1_ebtimeF_dtTrue_Exp_0>>>>>>>>>>>>>>>>>>>>>>>>>>
train 36456
val 5175
test 10444
iters: 100, epoch: 1 | loss: 0.6968859
speed: 0.8145s/iter; left time: 23051.8019s
iters: 200, epoch: 1 | loss: 0.6983635
speed: 0.6318s/iter; left time: 17817.0645s
Epoch: 1 cost time: 198.10548210144043
Epoch: 1, Steps: 284 | Train Loss: 0.7457502 Vali Loss: 0.5418562 Test Loss: 0.2217388
Validation loss decreased (inf --> 0.541856). Saving model ...
Updating learning rate to 0.0001
iters: 100, epoch: 2 | loss: 0.3526905
speed: 1.9244s/iter; left time: 53915.4483s
iters: 200, epoch: 2 | loss: 0.4079723
speed: 0.6331s/iter; left time: 17674.2574s

I found that set num_works to 0 is helpful.

[codeNiuMa]

look like you are using GPU,Use GPU: cuda:0, but its so slow!
if you are running on Windows, you need to change args.num_worker=0, it maybe helpful

[syrGitHub]

我也有类似的问题，我的问题是随着.sh文件里面的命令的运行，代码突然会变慢，请问您解决了吗？